The present invention relates to calibration techniques for film scanners, and more particularly to a technique using a calibration strip.
Digital Photofinishing provides the capability to improve images beyond what is possible in Optical Photofinishing. One improvement is increasing the gamma in the underexposed regions of customer negatives, to restore the final print to an appearance similar to that of a normally exposed negative. This improvement is called FUGC (Far Under Gamma Correction) and is described in U.S. Pat. No. 5,134,573 issued Jul. 28, 1992 to Goodwin, entitled xe2x80x9cMethod to Extend the Linear Range of Images Captured on Film.xe2x80x9d Another improvement is to alter distribution of pixels across the lightness range of the print in a way that provides improved shadow and highlight detail. A number of technologies have been created to perform this task both separately and in combination. FUGC and Contrast Normalization are described in U. S. application Ser. No. 09/086,333, now U.S. Pat. No. 6,233,069, filed May 28, 1998, by Buhr et al., entitled xe2x80x9cDigital Photofinishing System Including Film Under Exposure Gamma, Scene Balance, Contrast Normalization, and Image Sharpening Ditital Image Processingxe2x80x9d and U.S. Ser. No. 09/086,146, now U.S. Pat. No. 6,097,471, filed May 28, 1998, allowed Mar. 21, 2000, by Buhr et al., entitled xe2x80x9cDigital Photofinishing System Including Film Under-Exposure Gamma, Scene Balance, and Image Sharpening Digital Image Processing.xe2x80x9d Both of these technologies depend on the ability of the scanner to provide accurate printing density values for each pixel.
These technologies also require that the scene balance is estimated accurately and this is accomplished by the Single Channel Printing Algorithm (SCPA) which is described in U.S. Pat. No. 5,959,720, issued Sep. 28, 1999 to Kwon et al., entitled xe2x80x9cMethod for Color Balance Determination.xe2x80x9d
Both SCPA and FUGC depend especially on the scanner estimate of film minimum density (Dmin) expressed as printing density. The scanner observes areas outside of the image areas, where no exposure has taken place, and provides these values as measured Dmin. SCPA attempts to estimate a gray point for each frame in an order. It uses this measured Dmin to estimate the saturation of frames and selects frames to exclude from its calculation of gray. It also includes this measured Dmin as part of the formula for estimation of gray. So it can be seen that all of the tone scaling which is done by other algorithms, depends on the estimation of the scene gray point which in turn depends on the accuracy of measured Dmin provided by the scanner. FUGC increases the contrast in shadows by shifting the measured Dmin and applying a lookup table to increase the contrast of the shadows then shifting back to the original measured Dmin. As the contrast of shadows is increased, errors in measured Dmin are amplified in a way that can cause unnaturally colored shadows or dark subject areas in prints.
To assure that the improvements embodied in Digital Photofinishing are delivered faithfully and without artifacts or color errors, it is critical for the scanner to provide accurate printing density measurements of film Dmin. Film scanners in current Digital Photofinishing Systems are calibrated by measuring the scanner densities of the frames of a 17 patch calibration strip. These patches are exposed uniformly and include a series of 5 neutral patches and 12 colored patches. The intent in this original design is to present to the scanner, a range of densities representing 95% of the Large Area Transmission Densities of consumer color negatives. Currently used calibration strips do not contain a Dmin patch. Presumably this was omitted in the 17 patch calibration strip because it contains no pictorial information and was not important for optical printing algorithms.
For each of the 17 patches (i=1-17), the Red Aim Printing Density Ari is calculated. Ari=Red Reference Printing Density of the ith patch where                               Red          ⁢                      xe2x80x83                    ⁢          Reference          ⁢                      xe2x80x83                    ⁢          Printing          ⁢                      xe2x80x83                    ⁢          Density                =                              -            Log                    ⁢                                                    ∫                                  λ                  ⁢                                      xe2x80x83                                    ⁢                  1                                λ2                            ⁢                                                                    P                    λ                                    ·                                      T                    λ                                    ·                                      S                    λ                    r                                                  ⁢                                  xe2x80x83                                ⁢                                  ⅆ                  λ                                                                                    ∫                λ1                λ2                            ⁢                                                                    P                    λ                                    ·                                      S                    λ                    r                                                  ⁢                                  xe2x80x83                                ⁢                                  ⅆ                  λ                                                                                        (        1        )            
and
Srxcex=Kodak Edge 7 Paper Spectral Sensitivity for the red record;
P=Printer Light Source (average of Kodak subtractive printers)
T=Film Transmittance of the ith patch measured on spectrophotometer
xcex=Wavelength
The green and blue Aim Printing Densities are calculated in a similar fashion using the corresponding Kodak Edge 7 Paper Spectral Sensitivities for the green and blue records.
The measured red, green and blue scanner densities Sr, Sg and Sb of all 17 patches are linearly regressed against the aim red, green and blue densities of all the patches so as to minimize the errors in the summation:
Error Sum of Squares=xcexa3i{(Srixe2x88x92Ari)2+(Sgixe2x88x92Agi)2+(Sbixe2x88x92Abi)2}xe2x80x83xe2x80x83(2)
The resulting 3xc3x974 Calibration Matrix M provides a conversion from Scanner Densities S to Printing Densities P using Equation 3.
Pr=(m11xc3x97Sr+m12xc3x97Sg+m13xc3x97Sb+m14r)
Pg=(m21xc3x97Sr+m22xc3x97Sg+m23xc3x97Sb+m24r)
Pb=(m11xc3x97Sr+m32xc3x97Sg+m33xc3x97Sb+m34r)xe2x80x83xe2x80x83(3)
The range of densities over which accurate printing densities are provided by the scanners after calibration is limited by the range of densities in the 17 patches. Dmin values are determined by extrapolation from the calibrated range of densities on the calibration strip. This extrapolation can result in errors in measured Dmin, which as noted above will adversely affect the performance of digital image processing algorithms such as FUGC and SCPA.
There is a need therefore for an improved technique of scanner calibration that avoids this problem.
The need is met according to the present invention by providing a method of calibrating a scanner in a digital photofinishing system that includes the steps of: providing a calibration strip having a series of calibration patches including a plurality of neutral and colored patches, and including a Dmin patch; providing reference printing density values for each of the patches on the calibration strip; scanning the calibration strip in the scanner to produce scanner densities for each patch; and performing a regression on the scanner densities and the reference printing densities to produce a calibration matrix for converting from scanner density to printing density.
The calibration method with the strip having a Dmin patch embodied in this invention delivers improved Dmin accuracy without reducing accuracy in other densities. A further advantage is that the improved technique is compatible with the current technique, requiring only a software change in the program that drives the sensitometer to produce the calibration strip. No modification whatever is required in the scanners to use the improved strip. The method of the present invention employing the calibration strip with a Dmin patch produces enhanced image quality in the form of improved color consistency especially in enhanced prints from under-exposed negative.